1 /* 2 * xfrm6_policy.c: based on xfrm4_policy.c 3 * 4 * Authors: 5 * Mitsuru KANDA @USAGI 6 * Kazunori MIYAZAWA @USAGI 7 * Kunihiro Ishiguro <kunihiro@ipinfusion.com> 8 * IPv6 support 9 * YOSHIFUJI Hideaki 10 * Split up af-specific portion 11 * 12 */ 13 14 #include <linux/err.h> 15 #include <linux/kernel.h> 16 #include <linux/netdevice.h> 17 #include <net/addrconf.h> 18 #include <net/dst.h> 19 #include <net/xfrm.h> 20 #include <net/ip.h> 21 #include <net/ipv6.h> 22 #include <net/ip6_route.h> 23 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) 24 #include <net/mip6.h> 25 #endif 26 27 static struct xfrm_policy_afinfo xfrm6_policy_afinfo; 28 29 static struct dst_entry *xfrm6_dst_lookup(struct net *net, int tos, 30 xfrm_address_t *saddr, 31 xfrm_address_t *daddr) 32 { 33 struct flowi fl = {}; 34 struct dst_entry *dst; 35 int err; 36 37 memcpy(&fl.fl6_dst, daddr, sizeof(fl.fl6_dst)); 38 if (saddr) 39 memcpy(&fl.fl6_src, saddr, sizeof(fl.fl6_src)); 40 41 dst = ip6_route_output(net, NULL, &fl); 42 43 err = dst->error; 44 if (dst->error) { 45 dst_release(dst); 46 dst = ERR_PTR(err); 47 } 48 49 return dst; 50 } 51 52 static int xfrm6_get_saddr(struct net *net, 53 xfrm_address_t *saddr, xfrm_address_t *daddr) 54 { 55 struct dst_entry *dst; 56 struct net_device *dev; 57 58 dst = xfrm6_dst_lookup(net, 0, NULL, daddr); 59 if (IS_ERR(dst)) 60 return -EHOSTUNREACH; 61 62 dev = ip6_dst_idev(dst)->dev; 63 ipv6_dev_get_saddr(dev_net(dev), dev, 64 (struct in6_addr *)&daddr->a6, 0, 65 (struct in6_addr *)&saddr->a6); 66 dst_release(dst); 67 return 0; 68 } 69 70 static int xfrm6_get_tos(struct flowi *fl) 71 { 72 return 0; 73 } 74 75 static int xfrm6_init_path(struct xfrm_dst *path, struct dst_entry *dst, 76 int nfheader_len) 77 { 78 if (dst->ops->family == AF_INET6) { 79 struct rt6_info *rt = (struct rt6_info*)dst; 80 if (rt->rt6i_node) 81 path->path_cookie = rt->rt6i_node->fn_sernum; 82 } 83 84 path->u.rt6.rt6i_nfheader_len = nfheader_len; 85 86 return 0; 87 } 88 89 static int xfrm6_fill_dst(struct xfrm_dst *xdst, struct net_device *dev, 90 struct flowi *fl) 91 { 92 struct rt6_info *rt = (struct rt6_info*)xdst->route; 93 94 xdst->u.dst.dev = dev; 95 dev_hold(dev); 96 97 xdst->u.rt6.rt6i_idev = in6_dev_get(dev); 98 if (!xdst->u.rt6.rt6i_idev) 99 return -ENODEV; 100 101 /* Sheit... I remember I did this right. Apparently, 102 * it was magically lost, so this code needs audit */ 103 xdst->u.rt6.rt6i_flags = rt->rt6i_flags & (RTF_ANYCAST | 104 RTF_LOCAL); 105 xdst->u.rt6.rt6i_metric = rt->rt6i_metric; 106 xdst->u.rt6.rt6i_node = rt->rt6i_node; 107 if (rt->rt6i_node) 108 xdst->route_cookie = rt->rt6i_node->fn_sernum; 109 xdst->u.rt6.rt6i_gateway = rt->rt6i_gateway; 110 xdst->u.rt6.rt6i_dst = rt->rt6i_dst; 111 xdst->u.rt6.rt6i_src = rt->rt6i_src; 112 113 return 0; 114 } 115 116 static inline void 117 _decode_session6(struct sk_buff *skb, struct flowi *fl, int reverse) 118 { 119 int onlyproto = 0; 120 u16 offset = skb_network_header_len(skb); 121 struct ipv6hdr *hdr = ipv6_hdr(skb); 122 struct ipv6_opt_hdr *exthdr; 123 const unsigned char *nh = skb_network_header(skb); 124 u8 nexthdr = nh[IP6CB(skb)->nhoff]; 125 126 memset(fl, 0, sizeof(struct flowi)); 127 ipv6_addr_copy(&fl->fl6_dst, reverse ? &hdr->saddr : &hdr->daddr); 128 ipv6_addr_copy(&fl->fl6_src, reverse ? &hdr->daddr : &hdr->saddr); 129 130 while (nh + offset + 1 < skb->data || 131 pskb_may_pull(skb, nh + offset + 1 - skb->data)) { 132 nh = skb_network_header(skb); 133 exthdr = (struct ipv6_opt_hdr *)(nh + offset); 134 135 switch (nexthdr) { 136 case NEXTHDR_FRAGMENT: 137 onlyproto = 1; 138 case NEXTHDR_ROUTING: 139 case NEXTHDR_HOP: 140 case NEXTHDR_DEST: 141 offset += ipv6_optlen(exthdr); 142 nexthdr = exthdr->nexthdr; 143 exthdr = (struct ipv6_opt_hdr *)(nh + offset); 144 break; 145 146 case IPPROTO_UDP: 147 case IPPROTO_UDPLITE: 148 case IPPROTO_TCP: 149 case IPPROTO_SCTP: 150 case IPPROTO_DCCP: 151 if (!onlyproto && (nh + offset + 4 < skb->data || 152 pskb_may_pull(skb, nh + offset + 4 - skb->data))) { 153 __be16 *ports = (__be16 *)exthdr; 154 155 fl->fl_ip_sport = ports[!!reverse]; 156 fl->fl_ip_dport = ports[!reverse]; 157 } 158 fl->proto = nexthdr; 159 return; 160 161 case IPPROTO_ICMPV6: 162 if (!onlyproto && pskb_may_pull(skb, nh + offset + 2 - skb->data)) { 163 u8 *icmp = (u8 *)exthdr; 164 165 fl->fl_icmp_type = icmp[0]; 166 fl->fl_icmp_code = icmp[1]; 167 } 168 fl->proto = nexthdr; 169 return; 170 171 #if defined(CONFIG_IPV6_MIP6) || defined(CONFIG_IPV6_MIP6_MODULE) 172 case IPPROTO_MH: 173 if (!onlyproto && pskb_may_pull(skb, nh + offset + 3 - skb->data)) { 174 struct ip6_mh *mh; 175 mh = (struct ip6_mh *)exthdr; 176 177 fl->fl_mh_type = mh->ip6mh_type; 178 } 179 fl->proto = nexthdr; 180 return; 181 #endif 182 183 /* XXX Why are there these headers? */ 184 case IPPROTO_AH: 185 case IPPROTO_ESP: 186 case IPPROTO_COMP: 187 default: 188 fl->fl_ipsec_spi = 0; 189 fl->proto = nexthdr; 190 return; 191 } 192 } 193 } 194 195 static inline int xfrm6_garbage_collect(struct dst_ops *ops) 196 { 197 struct net *net = container_of(ops, struct net, xfrm.xfrm6_dst_ops); 198 199 xfrm6_policy_afinfo.garbage_collect(net); 200 return (atomic_read(&ops->entries) > ops->gc_thresh * 2); 201 } 202 203 static void xfrm6_update_pmtu(struct dst_entry *dst, u32 mtu) 204 { 205 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 206 struct dst_entry *path = xdst->route; 207 208 path->ops->update_pmtu(path, mtu); 209 } 210 211 static void xfrm6_dst_destroy(struct dst_entry *dst) 212 { 213 struct xfrm_dst *xdst = (struct xfrm_dst *)dst; 214 215 if (likely(xdst->u.rt6.rt6i_idev)) 216 in6_dev_put(xdst->u.rt6.rt6i_idev); 217 xfrm_dst_destroy(xdst); 218 } 219 220 static void xfrm6_dst_ifdown(struct dst_entry *dst, struct net_device *dev, 221 int unregister) 222 { 223 struct xfrm_dst *xdst; 224 225 if (!unregister) 226 return; 227 228 xdst = (struct xfrm_dst *)dst; 229 if (xdst->u.rt6.rt6i_idev->dev == dev) { 230 struct inet6_dev *loopback_idev = 231 in6_dev_get(dev_net(dev)->loopback_dev); 232 BUG_ON(!loopback_idev); 233 234 do { 235 in6_dev_put(xdst->u.rt6.rt6i_idev); 236 xdst->u.rt6.rt6i_idev = loopback_idev; 237 in6_dev_hold(loopback_idev); 238 xdst = (struct xfrm_dst *)xdst->u.dst.child; 239 } while (xdst->u.dst.xfrm); 240 241 __in6_dev_put(loopback_idev); 242 } 243 244 xfrm_dst_ifdown(dst, dev); 245 } 246 247 static struct dst_ops xfrm6_dst_ops = { 248 .family = AF_INET6, 249 .protocol = cpu_to_be16(ETH_P_IPV6), 250 .gc = xfrm6_garbage_collect, 251 .update_pmtu = xfrm6_update_pmtu, 252 .destroy = xfrm6_dst_destroy, 253 .ifdown = xfrm6_dst_ifdown, 254 .local_out = __ip6_local_out, 255 .gc_thresh = 1024, 256 .entries = ATOMIC_INIT(0), 257 }; 258 259 static struct xfrm_policy_afinfo xfrm6_policy_afinfo = { 260 .family = AF_INET6, 261 .dst_ops = &xfrm6_dst_ops, 262 .dst_lookup = xfrm6_dst_lookup, 263 .get_saddr = xfrm6_get_saddr, 264 .decode_session = _decode_session6, 265 .get_tos = xfrm6_get_tos, 266 .init_path = xfrm6_init_path, 267 .fill_dst = xfrm6_fill_dst, 268 }; 269 270 static int __init xfrm6_policy_init(void) 271 { 272 return xfrm_policy_register_afinfo(&xfrm6_policy_afinfo); 273 } 274 275 static void xfrm6_policy_fini(void) 276 { 277 xfrm_policy_unregister_afinfo(&xfrm6_policy_afinfo); 278 } 279 280 #ifdef CONFIG_SYSCTL 281 static struct ctl_table xfrm6_policy_table[] = { 282 { 283 .procname = "xfrm6_gc_thresh", 284 .data = &init_net.xfrm.xfrm6_dst_ops.gc_thresh, 285 .maxlen = sizeof(int), 286 .mode = 0644, 287 .proc_handler = proc_dointvec, 288 }, 289 { } 290 }; 291 292 static struct ctl_table_header *sysctl_hdr; 293 #endif 294 295 int __init xfrm6_init(void) 296 { 297 int ret; 298 unsigned int gc_thresh; 299 300 /* 301 * We need a good default value for the xfrm6 gc threshold. 302 * In ipv4 we set it to the route hash table size * 8, which 303 * is half the size of the maximaum route cache for ipv4. It 304 * would be good to do the same thing for v6, except the table is 305 * constructed differently here. Here each table for a net namespace 306 * can have FIB_TABLE_HASHSZ entries, so lets go with the same 307 * computation that we used for ipv4 here. Also, lets keep the initial 308 * gc_thresh to a minimum of 1024, since, the ipv6 route cache defaults 309 * to that as a minimum as well 310 */ 311 gc_thresh = FIB6_TABLE_HASHSZ * 8; 312 xfrm6_dst_ops.gc_thresh = (gc_thresh < 1024) ? 1024 : gc_thresh; 313 314 ret = xfrm6_policy_init(); 315 if (ret) 316 goto out; 317 318 ret = xfrm6_state_init(); 319 if (ret) 320 goto out_policy; 321 322 #ifdef CONFIG_SYSCTL 323 sysctl_hdr = register_net_sysctl_table(&init_net, net_ipv6_ctl_path, 324 xfrm6_policy_table); 325 #endif 326 out: 327 return ret; 328 out_policy: 329 xfrm6_policy_fini(); 330 goto out; 331 } 332 333 void xfrm6_fini(void) 334 { 335 #ifdef CONFIG_SYSCTL 336 if (sysctl_hdr) 337 unregister_net_sysctl_table(sysctl_hdr); 338 #endif 339 //xfrm6_input_fini(); 340 xfrm6_policy_fini(); 341 xfrm6_state_fini(); 342 } 343